AutoML Pipeline Predicts Emerging Trends from Text Data

A new research paper introduces a fully automated machine learning pipeline for predicting emerging trends from textual datasets with temporal attributes. The framework, detailed on arXiv, combines AutoClustering, AutoTopicModeling, and AutoTrendAnalysis. AutoClustering uses meta-learning to select the optimal clustering algorithm. AutoTopicModeling employs successive halving to choose among LDA, LSA, BERTopic, or NMF based on coherence scores. For trend forecasting, it evaluates models including Facebook Prophet and ARIMA. The system is designed to be scalable and adaptable for businesses, researchers, and policymakers.